Ahrefs builds its keyword difficulty score by counting how many websites link to each of the top 10 ranking pages.
Keysearch does something broader. Its algorithm weighs more than a dozen on-page and off-page factors, including backlinks, referring domains, the ranking strength of each main domain, and how well optimized the pages already sitting there happen to be.
Both dials run from 0 to 100, which is exactly why people trust them interchangeably. They should not. This is the part that gets lost in the argument over which keyword tool to trust, because both sides assume the two numbers are answering the same question when one is counting links and the other is grading pages.
Ten points at the bottom of the range might mean a handful of extra referring domains. Ten points at the top can mean hundreds. So a keyword sitting at 45 is not three times harder than one sitting at 15, whatever your instincts say.
Keysearch runs a color-banded system instead, moving from very easy through fairly easy, easy to moderate, moderate, fairly difficult, and very difficult. It does not publish the exact numeric cutoffs for those bands, and it says outright that the ranges may shift slightly when the team thinks they should.
Semrush draws its lines somewhere else again, using the median referring domains to the ranking URLs, the median ratio of dofollow to nofollow links, the median authority score of the ranking domains, and the qualities of the result page itself. Its published bands call anything from 30 to 49 possible rather than easy.
So the number you treat as your cutoff is not a fact about the keyword. It belongs to whichever tool you learned on. Someone trained on Keysearch will call a score of 29 a green light, while a Semrush user reads the same figure as merely easy, and a Keysearch veteran might already be reaching for the next idea.
It goes further. Ahrefs recommends checking the results page by hand if you actually want to gauge your chances, which is an unusual thing for a company to say about its own headline metric.
Keysearch is equally candid. It notes that difficulty "can vary wildly depending on which SEO tools you are using", and it describes its own algorithm as consistently tweaked whenever the team spots room to improve.
That second admission deserves more attention than it gets. If the formula keeps changing, then a score you saved in a spreadsheet eighteen months ago came out of a slightly different machine than today's. Comparing the two tells you very little about the keyword and quite a lot about the software.
Older cached scores are not hidden from you. They appear in a slightly dulled color rather than the usual full color, which is the only signal you get that a number was calculated at some point in the past rather than a moment ago.
There is a second layer to this. A setting called Bulk Check Credit Saver is switched on by default, and it skips cached scores entirely during bulk difficulty checks so that you burn fewer searches. Handy, until you forget it is running.
Turn it off, and the tool refreshes more of your list, though it still skips anything checked very recently. Either way, a bulk export is a mixture of fresh numbers and old ones wearing slightly different shades.
Two people can therefore pull the same keyword on the same afternoon and see different scores, purely because one of them got served a cached value and the other triggered a live check. Neither is wrong. Neither is current, either.
The practical fix costs nothing. When a keyword is about to become a page you will spend real hours writing, check that one individually and watch for the full color, rather than trusting the row it occupied in a bulk run last week.
Keysearch does something broader. Its algorithm weighs more than a dozen on-page and off-page factors, including backlinks, referring domains, the ranking strength of each main domain, and how well optimized the pages already sitting there happen to be.
Both dials run from 0 to 100, which is exactly why people trust them interchangeably. They should not. This is the part that gets lost in the argument over which keyword tool to trust, because both sides assume the two numbers are answering the same question when one is counting links and the other is grading pages.
The scales are not measuring the same thing
Ahrefs plots its score on a logarithmic scale. That single detail explains most of the confusion, because the distance between 10 and 20 is nothing like the distance between 70 and 80.Ten points at the bottom of the range might mean a handful of extra referring domains. Ten points at the top can mean hundreds. So a keyword sitting at 45 is not three times harder than one sitting at 15, whatever your instincts say.
Keysearch runs a color-banded system instead, moving from very easy through fairly easy, easy to moderate, moderate, fairly difficult, and very difficult. It does not publish the exact numeric cutoffs for those bands, and it says outright that the ranges may shift slightly when the team thinks they should.
Semrush draws its lines somewhere else again, using the median referring domains to the ranking URLs, the median ratio of dofollow to nofollow links, the median authority score of the ranking domains, and the qualities of the result page itself. Its published bands call anything from 30 to 49 possible rather than easy.
So the number you treat as your cutoff is not a fact about the keyword. It belongs to whichever tool you learned on. Someone trained on Keysearch will call a score of 29 a green light, while a Semrush user reads the same figure as merely easy, and a Keysearch veteran might already be reaching for the next idea.
Both vendors quietly admit the number is soft
Ahrefs says so in its own glossary. The score is "always only an estimation" because Google does not disclose all of its ranking factors, and the company adds that different tools take different approaches to the same calculation.It goes further. Ahrefs recommends checking the results page by hand if you actually want to gauge your chances, which is an unusual thing for a company to say about its own headline metric.
Keysearch is equally candid. It notes that difficulty "can vary wildly depending on which SEO tools you are using", and it describes its own algorithm as consistently tweaked whenever the team spots room to improve.
That second admission deserves more attention than it gets. If the formula keeps changing, then a score you saved in a spreadsheet eighteen months ago came out of a slightly different machine than today's. Comparing the two tells you very little about the keyword and quite a lot about the software.
The score you are reading may be weeks old
Keysearch does not check every keyword live. It blends real-time scoring with a cached database, and its real-time cache holds scores for anywhere between 7 and 30 days depending on your location.Older cached scores are not hidden from you. They appear in a slightly dulled color rather than the usual full color, which is the only signal you get that a number was calculated at some point in the past rather than a moment ago.
There is a second layer to this. A setting called Bulk Check Credit Saver is switched on by default, and it skips cached scores entirely during bulk difficulty checks so that you burn fewer searches. Handy, until you forget it is running.
Turn it off, and the tool refreshes more of your list, though it still skips anything checked very recently. Either way, a bulk export is a mixture of fresh numbers and old ones wearing slightly different shades.
Two people can therefore pull the same keyword on the same afternoon and see different scores, purely because one of them got served a cached value and the other triggered a live check. Neither is wrong. Neither is current, either.
The practical fix costs nothing. When a keyword is about to become a page you will spend real hours writing, check that one individually and watch for the full color, rather than trusting the row it occupied in a bulk run last week.
